create function on image.py and call it on app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,11 @@
|
|
|
|
1 |
import streamlit as st
|
2 |
|
3 |
-
import os
|
4 |
x = st.slider('Select a value')
|
5 |
st.write(x, 'squared is', x * x)
|
6 |
|
7 |
-
|
|
|
|
|
|
|
|
|
|
1 |
+
from image import *
|
2 |
import streamlit as st
|
3 |
|
|
|
4 |
x = st.slider('Select a value')
|
5 |
st.write(x, 'squared is', x * x)
|
6 |
|
7 |
+
|
8 |
+
url = "https://i.imgur.com/qs0CxjE_d.webp?maxwidth=760&fidelity=grand"
|
9 |
+
text = "What is the two color's car ?"
|
10 |
+
|
11 |
+
st.write(image(url, text))
|
image.py
CHANGED
@@ -1,24 +1,25 @@
|
|
1 |
from transformers import ViltProcessor, ViltForQuestionAnswering
|
2 |
-
import streamlit as st
|
3 |
import requests
|
4 |
from PIL import Image
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
# prepare
|
7 |
-
|
8 |
-
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
-
processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
12 |
-
model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
13 |
|
14 |
-
# prepare inputs
|
15 |
-
encoding = processor(image, text, return_tensors="pt")
|
16 |
|
17 |
-
#
|
18 |
-
|
19 |
-
logits = outputs.logits
|
20 |
-
idx = logits.argmax(-1).item()
|
21 |
-
st.write("Predicted answer:", model.config.id2label[idx])
|
22 |
-
print("question asked:", text)
|
23 |
-
print("image link:", url)
|
24 |
-
print("Predicted answer:", model.config.id2label[idx])
|
|
|
1 |
from transformers import ViltProcessor, ViltForQuestionAnswering
|
|
|
2 |
import requests
|
3 |
from PIL import Image
|
4 |
+
def image(url, text):
|
5 |
+
image = Image.open(requests.get(url, stream=True).raw)
|
6 |
+
processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
7 |
+
model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
8 |
|
9 |
+
# prepare inputs
|
10 |
+
encoding = processor(image, text, return_tensors="pt")
|
11 |
+
|
12 |
+
# forward pass
|
13 |
+
outputs = model(**encoding)
|
14 |
+
logits = outputs.logits
|
15 |
+
idx = logits.argmax(-1).item()
|
16 |
+
print("question asked:", text)
|
17 |
+
print("image link:", url)
|
18 |
+
print("Predicted answer:", model.config.id2label[idx])
|
19 |
+
|
20 |
+
return "Predicted answer:", model.config.id2label[idx]
|
21 |
|
|
|
|
|
22 |
|
|
|
|
|
23 |
|
24 |
+
# prepare image + question
|
25 |
+
|
|
|
|
|
|
|
|
|
|
|
|